Skip to content
All library documents

Risk Control, Kelly Sizing, and Sharpe Ratio in Trading

Article FMZ forum · Author: 发明者量化-小小梦

Summary

The article uses gambling and trading stories to explain why survival and capital preservation come before pursuing returns. Roulette illustrates how a small house edge compounds with repeated play, while blackjack card counting shows how a positive edge can emerge when the composition of the remaining cards becomes favorable. The author connects this to investing by arguing that trades should be taken only when expected value is positive and that position size should reflect both the edge and the odds.

It presents the Kelly formula as a way to size bets, then uses Jesse Livermore’s large positions as a cautionary example of how leverage can turn skill into ruin. The article also introduces the Sharpe ratio and maximum drawdown to compare return against volatility and losses, including examples from hedge funds and U.S. equities. These measures are presented as decision aids, not guarantees: the Sharpe ratio relies on simplifying assumptions such as normally distributed returns, while market returns can have serial dependence and fat tails. The discussion is educational and does not establish that any strategy or fund will repeat past performance.

Key ideas

  • Preserving capital and surviving adverse periods are prerequisites for long-term trading success.
  • A positive expected edge is necessary before committing capital, but it does not determine a safe position size by itself.
  • The Kelly formula links position size to the estimated probability of winning and the payout odds.
  • Taking positions larger than the Kelly fraction can reduce long-run growth and raise the chance of ruin.
  • The Sharpe ratio compares excess expected return with return volatility, while maximum drawdown captures peak-to-trough loss.
  • Normal-return assumptions and historical performance limit how confidently these measures describe future risk.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.